Data Allocation for Scratch-pad Memory on Embedded Multi-core Systems to Achieve Predictability and to Reduce Wcet

Please enable javascript in your browser.

Borhani, Zahra | 2016

441 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 48289 (19)
University: Sharif University of Technology
Department: Computer Engineering
Advisor(s): Ejlali, Alireza
Abstract:
Many embedded multicore systems can benefit from a memory hierarchy to bridge the processor/memory speed gap. Embedded systems usually utilize a cacheless architecture to avoid the unpredictable behavior. Unlike caches, scratchpad memories (SPM) are software controlled and hence the execution time of applications on such systems can be accurately predicted. Scheduling the tasks of an embedded multicore application and partitioning the available SPM budget among these cores are two critical issues in such systems. Often, these are considered separately, such a decoupled approach may miss better quality schedules. In this paper, we present an integrated approach to task scheduling and SPM partitioning to further reduce the worst case execution time of embedded heterogeneous multicore systems. We use a genetic algorithm to allocate data on different memory units for multicore systems with heterogeneous cores. Results on benchmarks show the significant improvement from our proposed technique. Our simulation studies reveal that our solution provides the desired guarantees while achieving more than 11.6 percent in reducing WCET than previous methods
Keywords:
Scratch Pad Memory (SPM) ; Data Association ; Multicore Embedded System ; Heterogeneous Architecture

No TOC